Combining the Best Stock Selection Factors

ELI5/TLDR

Patrick O’Shaughnessy argues that the factor investing trade — buy cheap, buy trending, buy quality — has been so thoroughly mined that the easy alpha is gone. The fix, he says, is not to abandon factors but to use them more aggressively: smaller portfolios, weirder factors (high-conviction buybacks at cheap prices is his favourite), and a willingness to look nothing like the index. The asset management industry has gone the other way — broad, scalable, closet-indexed — because that is what gathers assets, not what beats markets.

The Full Story

The factor trade has been mined out

O’Shaughnessy opens with a chart of the five canonical factors from Ken French’s website — value (book-to-price), momentum, profitability, investment, size — and their rolling 10-year long/short spreads. For most of the last several decades, going long the cheapest decile and short the most expensive was, in his words, “one of the markets’ versions of a free lunch.” Then, sometime around the zero-interest-rate era, every line collapses toward zero.

The trend has vanished, killed by its discovery.

He attributes the quote to Benoit Mandelbrot, whose The Misbehavior of Markets he plugs hard. The mechanism is plain enough: factor papers get published, hundreds of billions chase them, the valuation gap between cheap and expensive narrows, and the spread disappears.

Goodhart’s Law eats price-to-book

Price-to-book is the case study. In 1993, when Fama and French canonised it, all the standard value metrics — book, earnings, sales, EBITDA — showed similar excess returns. P/B got picked because it had the lowest turnover, meaning lower trading costs. Since the paper, P/B has been by far the worst-performing value metric. Swap sales-to-price in for book-to-price over the last ten years, he says, and your returns are 100% better cumulatively. The “value is dead” narrative is really a “price-to-book is dead” narrative.

He walks through Goodhart’s Law via two colonial bounty stories: Vietnam paid for rat tails and got tailless rats released back to breed; India paid for dead cobras and got cobra farms. When a measure becomes a target, it stops being a good measure. Price-to-book became a target.

Momentum is real but excruciating

Momentum has held up better than value lately, but the way it fails is brutal. Coming out of 2009, the long/short momentum spread was negative roughly 500% in a year — long the defensives that held up in the crash, short the financials and cyclicals that then ripped 400-700% off the bottom. This is the discipline tax. Most of the dollar-weighted returns from factor strategies, he argues, have gone to the fund sponsors, not the investors, because clients pile in late and out early.

His pet factor: high-conviction buybacks at cheap prices

This is the part of the talk where O’Shaughnessy is selling something, and he says so directly. His firm’s edge is a factor he claims is hard to commoditise into an ETF because it does not scale: companies repurchasing more than 10% of their float in a year while trading in the cheapest decile.

The distribution is striking. Low-conviction buyback firms — Apple buying back $1 billion against a $1.6 trillion market cap — show no preference for cheap valuations. High-conviction firms (a company buying back 10% of itself) heavily over-index on cheap stocks. Returns over 30 years:

Net share issuers: roughly 1% below the market
Low-conviction buybacks: roughly 1% above
High-conviction buybacks: roughly 5% above, annualised

A note on signal hygiene: he uses the trailing 12 months of actual repurchase activity, not announcements. Studies find only 26% of announced buybacks ever happen.

How the factors combine

The framing he uses is value investing as the house in Vegas. Only 51-52% of value stocks beat the market in any given year — a blackjack-sized edge. But there has never been a 10 or 15-year window where deep-value baskets failed to beat the market. The distribution of growth stocks has the fatter right tail (the Teslas and Amazons), which is exactly what keeps people walking back to the table.

He combines four families: cheapness (an equal-weight blend of P/S, P/E, EV/EBITDA, P/FCF — not P/B), momentum, quality (a fundamental read of balance sheets and accounting), and shareholder yield (dividends plus buybacks net of issuance). The combination math matters. High shareholder yield at cheap prices earned 13.7% annualised; high shareholder yield at expensive prices was a notable underperformer. The factors interact.

The closing argument is structural. “Active share” measures how different a portfolio is from its benchmark — 0 means index, 100 means no overlap. The industry has drifted toward closet indexing because agency risk punishes managers for tracking error.

If I have a more active portfolio… if I strike one of those bad years, the odds of my being fired go up very quickly.

So managers build portfolios with 300 names and 30% active share. They are unlikely to get fired. They are also unlikely to beat the market by enough to justify their fees. His prescription is the opposite: 50-55 stocks, no industry hedging, willingness to look nothing like the S&P 500, and three-to-five-year stretches of underperformance baked in as the price of admission.

The capacity ceiling is the point, not a bug. A strategy this concentrated cannot absorb $30 billion. That is why the major ETF sponsors will not build it, which is why it might still work.

A revealing aside on the Russell 1000 Value

Exxon Mobil was the largest holding in the iShares Russell 1000 Value ETF (which held $25 billion at the time of the talk). Exxon sits in only the 50th percentile of the index by price-to-book. It is the largest weight because the index is fundamentally market-cap weighted with a value tilt — meaning size is doing more work than value. “Any exposure to the market cap factor dilutes your advantage.”

Key Takeaways

The big five factors (value, momentum, profitability, investment, size) have all seen their long/short spreads compress over the past decade. Most of this is investor crowding, not the death of the underlying ideas.
Price-to-book is broken; other value metrics work better. O’Shaughnessy uses an equal-weight blend of P/S, P/E, EV/EBITDA, and P/FCF.
High-conviction buybacks (>10% of float) at cheap prices delivered ~5% annualised excess returns over 30 years. Use 12-month trailing actual repurchases, not announcements (only 26% of announced buybacks happen).
Net issuers underperformed by ~1% annually; low-conviction buyback firms outperformed by ~1%. The conviction tier matters more than the buyback itself.
Factors interact: high shareholder yield + cheap price = 13.7% annualised. High shareholder yield + expensive price underperforms. Combining factors is non-additive.
Value beats the market only 51-52% of years. The edge compounds over 10-15 year windows. Equity-curve volatility is the cost.
Momentum’s worst drawdowns happen at sharp market reversals. The 2009 long/short momentum unwind was roughly -500% in a year.
Active share matters as potential, not prediction. Closet indexers cannot beat the market by enough to cover fees. 50-stock portfolios with no industry hedges are the structural opposite of smart-beta ETFs.
Capacity is a feature. A factor that can absorb $30 billion has already been bid away. The strategies he favours top out in the low billions.
Long/short implementations of these factors mostly fail in practice — the short side is expensive or impossible to borrow, and unhedged industry exposure can blow up the book in a single sector move.

Claude’s Take

The honest version of this talk is: factor investing worked until it didn’t, the industry productised it into closet-index ETFs, and the only edge left is to do the thing that does not scale. That is a defensible position, and the high-conviction buybacks insight is genuinely interesting — the conviction tiering is the kind of nuance that survives because it cannot be ETF-ified at $50 billion of AUM.

The thing worth flagging is that the live track records of value-quant funds across the past decade have generally been worse than the backtests in talks like this one suggested. AQR’s flagship long/short value vehicles, Research Affiliates’ fundamental indexing, GMO’s deep value mandates — most of them underperformed in the 2010s in ways that the slide-deck backtests did not warn about. O’Shaughnessy’s own large-cap and small-cap composites that he shows (240% and 263% cumulative since 2004) are not third-party verified in the slides, and the dispersion between best- and worst-case value managers over the same period is enormous. The structural critique of the industry is sharp. The implicit promise that his version of factor investing has cracked the code deserves more skepticism than the talk format allows.

Score 8/10. The Goodhart’s Law framing of price-to-book, the conviction-tiered buyback factor, and the active-share-as-capacity argument are all sticky ideas that survive a re-read. Lose a point for the soft promotion of his firm’s strategies, lose another for the live-vs-backtest gap that the format conveniently elides. Read Mandelbrot.